Peter Straka

Research Fellow in Statistics

[Peter Straka @strakaps]

@robinren X... or Twitter?

[Peter Straka @strakaps]

@roberto_boberto Who's CD?

[Peter Straka @strakaps]

@TerribleMaps #maga works!

[Peter Straka @strakaps]

@mit_ide @CODEConference Is there someone I can contact with a question about registration please?

[Peter Straka @strakaps]

4. PhD Intern, London (https://t.co/UkD7pjEqo1)

And if you'd like to learn more about what it's like to work at FB Core Data Science, just drop me a line, I'd be very happy to chat!

[Peter Straka @strakaps]

Happy Monday!
My team, Core Data Science at Facebook EMEA is hiring:

1. Research Scientist (Stats/ML), London (https://t.co/fiJO3svXkG)
2. Research Scientist (EconCS), London
(https://t.co/ljUkuKwNPG)
3. Research Scientist (Stats/ML), Tel Aviv (https://t.co/742aEL5imr)

[Peter Straka @strakaps]

@myGovau Even after hearing all options, I cannot make a choice (my pressing the dialpad doesn't do anything). Tried from my partner's phone as well.
After hearing the whole recording twice, I was connected to an operator; just giving you a heads up that your system ain't working proper

[Peter Straka @strakaps]

@myGovau The call went through ok, it's just that once the menu is read out by the computerized voice, ("press 1 for mygov, 2 for ...") then pressing the keys doesn't do anything.

[zeynep tufekci @zeynep]

Good morning in sad, jealous tweets. All schools in Taiwan have been open since February. They did not listen to China or WHO in January and immediately masked up and shut down travel. They also traced the hell out of their outbreaks. They were practically done by February.

[Peter Straka @strakaps]

@petrmatous @ANSNA_Au .... unless, possibly, you use both projections (https://t.co/Z4SBVvsnYM, ht @RobAckland)! But I'm not sure how this pans out with blockmodels, and besides I'd like to avoid working with the patient-projection matrix, which has ~10^6 nodes

[Peter Straka @strakaps]

Predicting extremes in bursty time series #sixwordstory #Mathematics #Statistics

[Christina Farr @chrissyfarr]

Google is now saying:
Let's keep the research/AI parts under one group
Let's put products that are ready to scale in another.

Why this matters: It's among the first tech companies figuring out a real strategy for health vs. groups doing random projects.

https://t.co/gCMw1KEUmN

[Peter Straka @strakaps]

So you have a #logisticregression baseline model, and are thinking of the next step up in complexity? Before you fit a #deeplearning model, try glinternet first! Here's a little tutorial: https://t.co/2I0Nb8uZM4

[Peter Straka @strakaps]

@jakevdp The easiest way to see this is that in any interval of length x/n, you'll have 0 points with probability (1-x/n)^n, which converges to exp(-x), the tail function of the exponential distribution...

[Peter Straka @strakaps]

For this #computationalphysics paper everything happened in @rstudio:
* algorithms running in #rstats
* versioning on GitHub https://t.co/AidAV9oepp
* typesetting using #rmarkdown and the rticles package (https://t.co/nu7ScYqCUP)

[Peter Straka @strakaps]

Which single algorithm solves
* subdiffusion
* tempered subdiffusion
* variable order subdiffusion
* distributed order subdiffusion
* inverse subordinators?
Answer: this one https://t.co/E1q9fKzz5E (extends @james_nichols DTRW)

[Peter Straka @strakaps]

Could solve this interface problem: https://t.co/ZL8mwmcCvZ or this variable-order problem: https://t.co/DL2VU9NOdO

[Jean-Philip Piquemal @jppiquem]

#compchem Phys. Rev. Lett. 120, 248101 (2018) - Gradual Crossover from Subdiffusion to Normal Diffusion: A Many-Body Effect in Protein Surface Water https://t.co/T8yyeQW3DS

[Scott Sisson @ABC_Research]

Handbook of Approximate Bayesian Computation released worldwide today! Filled with outstanding content from outstanding ABC researchers! Well done to everyone who contributed!

[Peter Straka @strakaps]

Hierarchical stochastic blockmodels (the bipartite version) are better at topic modelling than Latent Dirichlet Allocation: